How to retrain the model

[fengzhangyin]

@sweetpeach I want to retrain the model, but I don't know what to do. Hope to give some detailed instructions. Thanks.

[sweetpeach]

To train the model, you can follow instructions from here. Then the trained model will be used in ReCode during decoding step at test time.

[carlos-gemmell]

@sweetpeach I see the device set for Theano is CPU in train.sh. Since CPU training from scratch would be slow, could you describe your setup for GPU training please. I have tried setting device in train.sh to "gpu" or "cuda". But Theano complains since "cuda" is for the old GPU backend and "gpu" gives me errors with pymc3 (which is installed).

A gpu docker container or some package versions for the requirements would be very helpful. Thanks.

[yz1019117968]

Hi, Could you please share a trained model with us for debugging purposes? Because I cannot easily understand the logic of your testing stage, especially the retrieval of ngrams and their alignments. Thanks a lot!!